The AI development life cycle: from planning to decommissioning
Every governance duty in this guide belongs to a stage of an AI system's life, so knowing the stages tells you when each duty applies. The stages also repeat: monitoring feeds back into planning, which is why governance work does not end at launch.
Why this matters for the exam
Every governance duty in this guide belongs to a stage of an AI system's life, so knowing the stages tells you when each duty applies. The stages also repeat: monitoring feeds back into planning, which is why governance work does not end at launch.
What you need to know
I.A covered what AI is and why it needs governing; I.B covered the people and structures that govern it. This competency is about what governance requires at each moment of a system's life. It starts with the timeline itself: the life cycle, from planning to decommissioning. The topic that follows maps its policy areas straight onto these stages.
The life cycle is a loop
The course defines the AI development life cycle as "the iterative, structured process of moving from a problem or idea to an AI solution." It looks like ordinary software development with one addition: continuous, rigorous monitoring. An AI system is built on data, and it has to keep performing as that data and the world around it change. The stages therefore repeat. Findings from monitoring routinely feed back into planning, and the cycle starts again.
The seven stages, and what governance requires at each
| Stage | Governance requirement at this stage |
|---|---|
| 1. Planning and design | Define the problem AI will solve; consider the user group; consider whether an interpretable model is appropriate. |
| 2. Data collection and preparation | Ensure data is representative of the problem; prevent bias in data labeling. |
| 3. Model development (selection & training) | Build in explainability by design; maintain appropriate documentation. |
| 4. Model testing and evaluation | Test for bias and maintain fairness principles; ensure adequate user testing and representation. |
| 5. Deployment | Enable user feedback channels; put an incident and error reporting function in place. |
| 6. Monitoring and maintenance | Set a monitoring and reporting schedule; run regular quality checks; have an action plan to retrain or take the model offline. |
| 7. Decommissioning | Properly archive or destroy sensitive data and stand the system down to prevent safety, reputational and legal risk; document the process. |
Decommissioning is easy to overlook, and it is a full stage of its own, with its own documentation and risk management. A system that was retired carelessly can still cause safety, reputational and legal problems.
Domains III and IV walk through these stages in much more operational depth. Domain III centers on the building stages, Domain IV on choosing, deploying and running a system. Monitoring and decommissioning governance appear in both.
Next up: the policies that ride on that timeline. The Body of Knowledge names nine life-cycle policy areas, and the first of them gets a full treatment: use case assessment.
Remember
- The life cycle is iterative rather than linear: findings from monitoring feed back into planning, and the stages repeat.
- The seven stages in order: planning and design, data collection and preparation, model development, testing and evaluation, deployment, monitoring and maintenance, decommissioning.
- Decommissioning is a full stage with its own documentation and risk management.
- Domain III expands the building stages, Domain IV expands deployment and operation, and monitoring and decommissioning governance appear in both.
Practise this topic
Domain I is free in the app, including its practice questions and flashcards, with progress tracking and no card details.
Previous: Building buy-in: leadership, training, AI literacy and culture
Next: Policies across the AI life cycle, and use case assessment
Back to the AIGP study guide.
AI Governance Study is an independent study aid. It does not represent a government entity: it is not affiliated with, endorsed by or authorised by any government, government agency or regulatory authority, and it does not provide government services or legal advice. Laws and frameworks are described in our own words — the official texts are listed at official sources. It is also not affiliated with, endorsed by, or sponsored by the IAPP. The AIGP name is used only to identify the exam this material helps you prepare for.